Volunteers sought: try out tutorials for speech processing with documentation data

Hi all!

Some of you may know that I’m writing a bunch of tutorials on https://parledoct.ai/:

Parledoct (like ‘aqueduct’) is a collection of tutorials and supporting tools to assist in the deployment of speech processing pipelines for language documentation projects. These tutorials are aimed at those who have some basic familiarity with speech data (e.g. annotating in Praat/ELAN) and scripting (e.g. in Praat or R) and want to learn to directly use modern speech processing tools via the command-line interface, complementing projects such as Elpis which provides a user-friendly interface for similar tools.

I was hoping to get some folks to try out 3 tutorials and give feedback:

I’m looking to get feedback to refine these earlier ones to make them as beginner friendly as possible (for now defined as: has some Praat/R scripting experience but little to no Python and speech processing experience), since the later ones will build on them.

Any and all general thoughts also welcome!

Thanks!

5 Likes

Looking over it, a general thought is that some more screenshots to orient the user to the data structure (especially when pulling in the external Ihanzu data) would be useful for cases where they can’t get the download to work. In my case, I’ll be away from reliable internet for more than a month shortly, so wget might not work. But if I can see what files are supposed to be present in the data it’d help me arrange my local files that way.

Time permitting, I may give the docker env a spin and let you know how it works. I have too much Python experience to be an ideal tester for you, but I teach undergrads so I can probably anticipate some basic issues.

2 Likes

Thanks @faytak!

Looking over it, a general thought is that some more screenshots to orient the user to the data structure (especially when pulling in the external Ihanzu data) would be useful for cases where they can’t get the download to work. In my case, I’ll be away from reliable internet for more than a month shortly, so wget might not work. But if I can see what files are supposed to be present in the data it’d help me arrange my local files that way.

Good point. I may just make a standalone ‘Exploring the Ihanzu data’ also as a way to introduce the interface, Shell commands, etc.

Time permitting, I may give the docker env a spin and let you know how it works. I have too much Python experience to be an ideal tester for you, but I teach undergrads so I can probably anticipate some basic issues.

That’d be great. I just had some feedback from two people in that target demographic who had spaces in their local directory names and so changing directory via cd C:\Users\path with space failed, so very much on the lookout for issues like these (whether anticipated or encountered…)

1 Like

Just wanted to jump in to say for the record (you know this already, @fauxneticien!) that I have been trying out these tutorials too and they are all working well. Docker is new to me so it has been interesting in that regard as well.

Your “tree mode” of tutorials is a great idea, I hope it becomes a sequoia! :evergreen_tree:

1 Like

I haven’t looked at this closely, but I enjoyed your presentation today. If you feel it’s ready, you might propose a workshop at ICLDC7: http://ling.lll.hawaii.edu/sites/icldc/workshop-and-talk-story-proposals/

2 Likes

Thanks @SarahRMoeller! I see the deadline’s August 1. Hm I’ll keep this in mind as I write material and decide mid-July if it might ready for that deadline…

2 Likes

Hi @SarahRMoeller!

I think that call for workshops deserves its own topic, I’ll try to put one together if you don’t beat me to it :slight_smile:

The conference has been great, hoping to see a few more talks tomorrow and watch many of the recorded talks soon.

Cool! Will try this out. One thing that I think would be hugely helpful for those relatively unfamiliar to the terminal is a set of common errors and their solution. This comes both from teaching undergrads and getting persephone running. Of course, the number of errors is limitless, but the common mistakes are more constrained (e.g. “says it can’t find your file? check you’re in the right directory; check you spelled the name correctly”, etc - not specific to this app, but the things that trip up new users often aren’t.

3 Likes

Thanks @cbowern!

One thing that I think would be hugely helpful for those relatively unfamiliar to the terminal is a set of common errors and their solution.

I’m planning on implementing a ‘FEE’ (frequently encountered errors, in the style of an FAQ page), inspired by the one that exists for the An Introduction to Statistical Learning textbook here. It’s just a single page with a Disqus thread where people can post new questions for any tutorial. I’ll also curate/append the questions and solutions in a table-of-contents style list as they come in…

1 Like

Wonderful! There’s a fantastic thread where someone posted all the HTK errors they encountered and how they solved them, which saved me many many hours